39B-8

O. TOKUSOGLU and M. O. Balaban. University of Florida, FSHN Department, Gainesville, FL 32611

JUSTIFICATION: Current quality assessment of grapes relies on sensory inspections . These are subjective, time-consuming, expensive and not always reproducible. Electronic noses have potential in the food industry as quality control and screening tools. Color is also subjectively evaluated. Machine-vision can quantify non-uniform colors. Pattern recognition can be used for both e-nose and machine vision data towards objective classification of grapes.

OBJECTIVE: The objectives of this study were 1) to correlate e-nose readings with odor sensory evaluation of muscadine grapes stored at different temperatures for different times, 2) to correlate machine vision readings with color sensory evaluations.

METHODS: Black muscadine grapes (Vitis rotundifolia Noble) were stored at 1.8, 7.0 and 11.7 C for up to 15 days. Replicate samples of about 100g were used for e-nose (model 4000, EEV) and color readings, and sensory evaluations at days 1, 3, 6, 9, 12, 15. A trained sensory panel of 13 judged the samples. E-nose and color data were correlated with sensory data using Discriminant Function Analysis (DFA).

RESULTS: DFA analysis of e-nose data correlated with storage time at 1.8, 7.0 and 11.7 C gave correct classification rates of 100%, 100% and 98%, respectively. When classified by temperature at each day, the classification rates for days 1, 3, 6, 9, 12 and 15 were 100%, 100%, 87%, 93%, 97%, 100%, respectively. Classification of e-nose data by sensory readings gave classification rates of 100%, 97%,97% for 1.8, 7.0 and 11.7 C, respectively. Color analysis resulted in similar values.

SIGNIFICANCE: After extensive training, the e-nose can be used reliably to classify grape samples by odor, and these can be closely matched to sensory readings. Machine vision can objectively classify grape samples by color. DFA can be used as a pattern recognition technique for the identification and determination of food odor and color.